The present invention relates to an image processing device and, more specifically, to a scheme for correcting a DC offset of an image signal received from an image sensor to a reference value.
In electrophotographic copiers, image scanners, etc., an image signal received from an image sensor such as a CCD has a DC offset, which is different from one image sensor to another and varies over years and with temperature. Since a variation of the offset value causes a black reference level variation to deteriorate image quality, it is necessary to correct the DC offset to a reference value.
For example, Japanese Patent Application Examined Publication No. Sho. 60-28183 discloses a technique of correcting a DC offset to a reference value. In this technique, an offset level is detected by sampling and holding a dummy bit portion of an image signal. The detected offset level is compared with a reference value to judge which is larger, and a value corresponding to the offset level is obtained by integrating pulse outputs in accordance with a judgment result. A signal representing the value corresponding to the offset level is subtracted from an input signal to cancel the DC offset.
In the above conventional offset canceling technique, noise included in an image signal makes it difficult to correctly adjust the timing of the sample-and-hold operation to perform correct level detection. Since a time constant is set at a relatively large value, time is required for the circuit to reach a stable state, for instance, at the time of power on. Further, the control cannot be performed in a desired manner.